End of training
Browse files
README.md
CHANGED
|
@@ -2,12 +2,13 @@
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
|
|
|
|
|
|
| 5 |
- generated_from_trainer
|
| 6 |
base_model: TheBloke/Mistral-7B-Instruct-v0.1-GPTQ
|
| 7 |
model-index:
|
| 8 |
- name: mistral-finetuned-samsum
|
| 9 |
results: []
|
| 10 |
-
pipeline_tag: text-generation
|
| 11 |
---
|
| 12 |
|
| 13 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -29,38 +30,17 @@ More information needed
|
|
| 29 |
|
| 30 |
More information needed
|
| 31 |
|
| 32 |
-
|
| 33 |
## Training procedure
|
| 34 |
|
| 35 |
-
|
| 36 |
-
The following `bitsandbytes` quantization config was used during training:
|
| 37 |
-
- quant_method: gptq
|
| 38 |
-
- bits: 4
|
| 39 |
-
- tokenizer: None
|
| 40 |
-
- dataset: None
|
| 41 |
-
- group_size: 128
|
| 42 |
-
- damp_percent: 0.1
|
| 43 |
-
- desc_act: True
|
| 44 |
-
- sym: True
|
| 45 |
-
- true_sequential: True
|
| 46 |
-
- use_cuda_fp16: False
|
| 47 |
-
- model_seqlen: None
|
| 48 |
-
- block_name_to_quantize: None
|
| 49 |
-
- module_name_preceding_first_block: None
|
| 50 |
-
- batch_size: 1
|
| 51 |
-
- pad_token_id: None
|
| 52 |
-
- use_exllama: False
|
| 53 |
-
- max_input_length: None
|
| 54 |
-
- exllama_config: {'version': <ExllamaVersion.ONE: 1>}
|
| 55 |
-
- cache_block_outputs: True
|
| 56 |
-
|
| 57 |
### Training hyperparameters
|
| 58 |
|
| 59 |
The following hyperparameters were used during training:
|
| 60 |
- learning_rate: 0.0002
|
| 61 |
-
- train_batch_size:
|
| 62 |
- eval_batch_size: 8
|
| 63 |
- seed: 42
|
|
|
|
|
|
|
| 64 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 65 |
- lr_scheduler_type: cosine
|
| 66 |
- training_steps: 250
|
|
@@ -72,8 +52,8 @@ The following hyperparameters were used during training:
|
|
| 72 |
|
| 73 |
### Framework versions
|
| 74 |
|
| 75 |
-
- PEFT 0.7.
|
| 76 |
-
- Transformers 4.
|
| 77 |
- Pytorch 2.1.0+cu118
|
| 78 |
-
- Datasets 2.
|
| 79 |
- Tokenizers 0.15.0
|
|
|
|
| 2 |
license: apache-2.0
|
| 3 |
library_name: peft
|
| 4 |
tags:
|
| 5 |
+
- trl
|
| 6 |
+
- sft
|
| 7 |
- generated_from_trainer
|
| 8 |
base_model: TheBloke/Mistral-7B-Instruct-v0.1-GPTQ
|
| 9 |
model-index:
|
| 10 |
- name: mistral-finetuned-samsum
|
| 11 |
results: []
|
|
|
|
| 12 |
---
|
| 13 |
|
| 14 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 30 |
|
| 31 |
More information needed
|
| 32 |
|
|
|
|
| 33 |
## Training procedure
|
| 34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
### Training hyperparameters
|
| 36 |
|
| 37 |
The following hyperparameters were used during training:
|
| 38 |
- learning_rate: 0.0002
|
| 39 |
+
- train_batch_size: 6
|
| 40 |
- eval_batch_size: 8
|
| 41 |
- seed: 42
|
| 42 |
+
- gradient_accumulation_steps: 4
|
| 43 |
+
- total_train_batch_size: 24
|
| 44 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 45 |
- lr_scheduler_type: cosine
|
| 46 |
- training_steps: 250
|
|
|
|
| 52 |
|
| 53 |
### Framework versions
|
| 54 |
|
| 55 |
+
- PEFT 0.7.1
|
| 56 |
+
- Transformers 4.37.0.dev0
|
| 57 |
- Pytorch 2.1.0+cu118
|
| 58 |
+
- Datasets 2.16.0
|
| 59 |
- Tokenizers 0.15.0
|